The overlay's auto-save implementation arrives via a dynamic import that
races the first React commit. A hook whose mount effect ran before the
overlay chunk evaluated saw installedImpl === null and stayed idle for
the whole life of the tab: no debounced saves, no beforeunload flush,
with no visible symptom. Any tab that hard-loaded straight into the
editor and never remounted it silently lost every edit made after the
project was bound.
installAutoSaveImpl now wakes hooks that mounted before it ran, so the
implementation starts as soon as it exists. Swapping a live impl at
runtime remains unsupported.
The load effect listed the overlay's `settled` flag as a dependency. On
a direct /example/<pro-id> URL the load can begin the moment the
overlay registers the example — one microtask before the overlay's
import promise settles. The flip then re-fired the effect mid-load: the
cleanup cancelled it (setReady skipped), the re-run hit the loadedIdRef
guard and returned, and the page hung on "Loading example…" forever.
The effect now depends only on the example itself; the 404-vs-still-
loading decision moved into the render, where reading `settled` cancels
nothing. A load cancelled mid-flight also resets the loadedIdRef guard
so a genuine re-run reloads instead of early-returning.
setMicrophoneSource on the simulator shim mirrors addI2CDevice: a part
that produces audio hands one 16-bit sample per call to whatever bridge
implements the method (the in-browser JS-engine bridges); everywhere
else the optional chain makes it a silent mic. No board-kind checks in
components. Plus the flat reSpeaker Lite glyph for gallery cards.
Every control in the unified strip is now 28px tall: tb-btn (was 30),
Libraries (was 26), the C++/MicroPython select (was ~22), the board
selector, Serial/Scope, canvas icon buttons, zoom and Add (all were 32).
The explorer toggle moves inside the Code/Both/Circuit segmented group
as its first segment, and every active-segment blue is the Libraries
blue (var(--color-action-primary)) instead of the hardcoded #0e639c.
The editor zone's overlap floor becomes min-width: max-content — the
hardcoded 280px sat below the real content width, letting the console
button paint over the board selector when the docked chat narrowed the
bar. Now the zone can never shrink under its buttons; flex-wrap moves
the canvas controls to a second line instead. Also closes the gap
between the logo and the File menu (header gap 16->6, menubar
margin-left 14->2).
The common namespace was assembled with a shallow spread, so any
top-level section present in two locale files had the later file clobber
the earlier one WHOLESALE. Two live casualties:
- header: adding header.shareProject to common2 (yesterday's raw-string
cleanup) erased common.json's whole header — header.auth.* and
header.nav.* — so the account menu rendered "header.auth.signOut" as
its label. My regression, caught verifying the account-menu fixes.
- editor.share.*: common2's editor section has been clobbering
common.json's since forever — the Share modal's seven visibility
labels were silently lost in every locale. Pre-existing.
Locale files merge by key now (deepMerge, both the synchronous en init
and the dynamic per-locale loader), which ends the failure class instead
of the instance; shareProject also moves to common.json's header where
it belonged. Verified: all header.auth/nav and editor.share leaves
resolve again in all nine locales.
The bottom-left block read wrong: a globe crowding the account button for
no gain — language already lives in the menubar's Language menu for
everyone, and (for signed-in users) inside the account menu. The footer
now holds only the account button.
Width-aware, as requested: when the explorer is dragged narrow (footer
under 150px, via container query) the username hides and the avatar
alone identifies the account; with the explorer collapsed, the floating
fallback box shows just the circle.
Adds the OSS half of the crash fix: LocaleSync listens for
'velxio-locale-switch' window events and performs the locale change as a
normal SPA navigation. The pro account menu is injected into its own
React root OUTSIDE the Router, so it cannot navigate itself — its
language rows dispatch this event instead.
The explorer toggle tooltips (Hide/Show file explorer) and the header's
Share-project title were raw English — now keyed and translated in all
nine locales, same common2.json home as the menubar strings.
Every menu we added lived on t() English fallbacks — the keys did not
exist in ANY locale file, so a Spanish or Chinese user got an English
menubar. 29 keys (editor.menu.* + the toolbar labels the File menu
reuses) now exist in en/es/pt-br/it/fr/zh-cn/de/ja/ru.
Two entries needed their own leaf keys: editor.toolbar.compile and .run
are nested OBJECTS (title/options/... sub-keys), so t() on them cannot
resolve to a string — the menu now uses editor.menu.compile / .run,
which also avoids double shortcuts ("Compile (Ctrl+B)" plus the menu's
own Ctrl+B column).
Verified mechanically: all 38 keys the menubar references resolve to
strings in all nine locales.
The menubar now matches the desktop app's native menu set: File, Edit,
View, Language, Help.
View collects what the user asked to reach from a menu: Compile (Ctrl+B),
Run, Stop, Reset up top; the panel toggles — File Explorer, Output
Console, Serial Monitor, Oscilloscope/Logic Analyzer — in the middle
(store-backed ones render a live check, like a real desktop menu); and
the canvas view actions (center, zoom) move here from Edit, which goes
back to being undo/redo only, as menus have always worked.
Language lists the nine locales with the current one checked, switching
through the same switchLocale path the header globe uses — so language is
reachable from the menubar regardless of the account state, on top of
living in the signed-in account menu.
Run/Stop/Compile/Reset and the console toggle register through the
editorCommands seam from EditorToolbar (their handlers close over its
state); the explorer toggle from EditorPage; serial and scope call their
stores directly, same as undo/redo.
The floating corner box sat ON TOP of the file tree — create enough files
and the last rows scrolled underneath it. The account + language block is
now the explorer panel's footer (VS Code style): the tree scrolls above
it, always. When the explorer is collapsed the block falls back to the
small fixed corner box, so account and language stay reachable.
The standalone language globe becomes a logged-out affordance: once the
pro auth slot renders its signed-in marker (data-auth-user), a :has()
rule hides the globe — signed-in users change language inside the
account menu, where the pro overlay now hosts it.
The language menu and the account menu were written for the top bar and
open downward; from the bottom-left corner box that means straight off
the bottom of the viewport. The language menu flips via a scoped CSS
rule; the account menu (pro) now anchors bottom-up whenever its trigger
sits in the lower half of the screen.
The header's backdrop-filter makes it a containing block for fixed
descendants, so the corner box's bottom:8px resolved against the 45px
header — measured at y=-4, sitting on the logo. Portaled to <body>, where
fixed means the viewport. The auth slot keeps working: pro's injector
finds it by attribute wherever it lives.
IDE-style: a small fixed box in the explorer's quiet bottom corner holds
the language switcher and the account button (pro keeps injecting it into
the same data-velxio-slot, just relocated). The autosave dot rides next
to the menus, and the header's Share button is gone — File > Share/Embed
already covers it.
The point is width: emptying the header's right side gives the toolbar
the last ~150px it needed, so the single-row layout now holds down to
~1400px WITH the chat docked — which is exactly the reported 1440x900
case. Tight-fallback thresholds drop accordingly (1400/1060/1040).
Desktop-editor variant only; marketing pages and mobile keep their
header untouched.
At 1440x900 with the chat docked, the merged header hit an ugly in-between:
the wrapped canvas controls floated as an orphan group inside a tall
transparent header, next to a dead gap before the language/user controls.
Reported by the user as "se ve raro" — wider screens looked fine.
Below the width where one row genuinely fits (thresholds account for the
docked chat's 380px / collapsed 36px reservation), the strip now falls
back to being its own full-width second bar with its old background and
border — the exact familiar pre-merge layout. Wide screens keep the
single 44px row; tight screens get the classic two bars; the broken-
looking limbo between them no longer exists.
The toolbar's overflow button held four leftovers (Share/Embed, Sync to
GitHub, Upload firmware, Record simulation). They move into the File
menu, which is where a session-frequency action belongs, and the strip
loses one more button. The PRO pill travels with them — same style the
overflow used (now .emb-pro) — and the two File items that were always
premium (BOM, schematic image) finally show it too: users should know an
item is premium BEFORE clicking, not via a surprise upgrade prompt.
The pro actions keep firing the same window events the overflow items
fired (share/github-sync/record prompts), so the overlay's listeners are
untouched and OSS builds keep their silent no-op. Dead overflow CSS
removed with the button.
With the marketing nav gone from the editor header, its middle was ~700px
of dead space at 1280 while the toolbar occupied a whole second 38px bar.
The unified toolbar strip now fills that middle through a new AppHeader
editorToolbar slot: one 44px row where there used to be 44+38 — a full
row returned to the code and the canvas.
The strip keeps its own class names, so everything that keyed on them
keeps working untouched: the container queries, the docked-chat
padding-right, and the internal flex-wrap. When it truly cannot fit, the
strip wraps and the header grows (height: auto on the modifier class)
instead of clipping or overlapping; brand and the right-side controls
stay pinned to the first line. Inside the header the strip drops its own
background and border so it reads as one bar, not a box within a bar.
Mobile keeps the previous layout (no strip; the mobile tab bar remains).
The open-source build now ships exactly its product surface: /editor, the
examples gallery (/examples, /examples/:id) and the per-example editor
(/example/:id). Landing, about, pricing, docs, the 14 keyword-targeted
simulator landings and the v2/v2.5/v3 showcases move to the private
overlay, registered through the same registerProRoutes seam that already
carries login/admin/classroom.
Root behaves per build: an overlay that registers an index route claims
'/' (velxio.dev keeps its landing); otherwise '/' redirects to /editor.
The redirect waits for the overlay import to settle — same contract as
markProExamplesSettled — so a velxio.dev visitor is never bounced into
the editor because the landing was 300ms away from registering.
Prerender and sitemap follow the split: entry-server pulls the marketing
page map from '@pro/pages/marketing' behind a VITE_PRO_BUILD-gated dynamic
import (the proven main.tsx pattern, with an OSS stub for tsc), and
generate-sitemap lists only served routes in OSS builds (2 URLs) while
pro builds stay byte-identical — verified against a pre-migration
baseline: sitemap (37 URLs) and prerendered /, /about, /docs,
/arduino-simulator, /esp32-simulator, /v3 all identical modulo hashed
asset names. OSS prerender drops exactly the 32 marketing pages (348→316).
The OSS header slims down to match: Editor, Examples, GitHub, Discord —
the marketing links only render in pro builds, where their routes exist.
The editor Help menu links those pages absolutely (velxio.dev) in OSS,
exactly like the desktop app's Help menu.
Mismo contenido y orden que el Help del menu nativo de escritorio
(pro/desktop menu.rs): Documentation, Examples, Pricing | Velxio Home,
Blog, About Velxio | Discord Community, GitHub Repository. Los enlaces que
el header del editor dejo de mostrar como nav de marketing recuperan aqui
un sitio ordenado, y todo abre en pestana nueva para que el editor (y el
trabajo sin guardar) se quede donde esta — el mismo criterio que la app
nativa, que abre el navegador del sistema.
Reportado en /example/pi5-opencv-vision: al recargar aparecia un instante
el 404 con el header completo (menus incluidos) y despues cargaba el
editor. La carrera: los ejemplos pro se registran cuando aterriza el
import dinamico del overlay, y la pagina resolvia la galeria ANTES,
concluia "no existe", pintaba el 404 y al llegar el registro re-renderizaba
al editor.
"No esta en la galeria" y "no esta TODAVIA" son respuestas distintas, asi
que el registro gana una senal de asentado: main.tsx la activa cuando el
import del overlay resuelve (o falla — finally), e inmediatamente en la
build OSS, donde no viene overlay y un 404 debe ser instantaneo. La pagina
se queda en "Loading example..." hasta que el registro asienta y solo
entonces un id ausente es de verdad un 404.
markProExamplesSettled es idempotente y solo notifica una vez, con test —
el sintoma de romper eso seria una tormenta de re-renders por hot-reload
del overlay.
Con la ventana estrecha y el chat acoplado, a la fila unificada le faltan
~60px y flexbox se los quitaba al toggle Code/Both/Circuit: de 101px a 32,
con dos de sus tres botones recortados a esquirlas inalcanzables bajo el
overflow hidden. Medido en staging a 1024px.
El toggle deja de encogerse (flex-shrink 0) y la fila hace wrap: cuando de
verdad no cabe, los controles del canvas bajan a una segunda linea corta.
Dos filas breves ganan a botones invisibles.
La queja: demasiados botones en una sola fila, y en pantallas pequenas se
solapan — medido, no impresion: a 1024px Code/Both/Circuit pisaban a
Compile/Run/Stop por 9-20px y Add pisaba el chat por 29px, con el idioma y
el usuario cortados fuera del viewport.
Tres piezas:
1. En el editor, el nav de marketing (Home/Docs/Pricing/...) sobra: ya
estas dentro, y cuesta exactamente el ancho que le falta a la toolbar.
AppHeader gana la variante editorMenu — mismo mecanismo que ya usa la
build de escritorio (VITE_DESKTOP) — que oculta el nav y pinta un menu
File/Edit junto al logo. El logo sigue llevando a inicio; idioma,
autoguardado, Share y usuario se quedan.
2. File/Edit al estilo escritorio: lo que se usa cada minuto sigue siendo
boton; lo que se usa unas veces por sesion va al menu. File: nuevo
workspace, nuevo fichero, abrir, guardar, importar, exportar, BOM,
imagen del esquema. Edit: undo/redo (con su estado real del historial),
centrar vista, zoom. Las acciones viven en cierres de cuatro
componentes distintos, asi que hay un registro id->handler
(lib/editorCommands, mismo patron de seam que registerProExamples o
registerBoardBuiltins): cada dueno registra al montar y el menu invoca
por id; un item sin dueno montado se pinta deshabilitado, que ademas es
la verdad. El registro sobrevive al remontaje StrictMode (el cleanup
viejo no borra al sucesor) y tiene test de eso.
3. La toolbar adelgaza: fuera Undo/Redo (Ctrl+Z/Y siguen), fuera los
botones inline de Import/Export y sus gemelos responsivos del More
(~70px recuperados); el More queda solo con los extras pro (firmware,
GitHub, share, grabacion). El CSS responsivo de los botones retirados
se va con ellos.
Suite en verde (2300) y build del overlay verificado desde este arbol.
Tres cambios que van juntos porque atacan la misma queja: "anado un
elemento y a veces ni veo donde se anadio".
1. Donde cae. Ya se anclaba a la esquina visible, pero la cascada que evita
que se apilen iba indexada por components.length: seguia avanzando
aunque movieras o borraras piezas, asi que las caidas se alejaban cada
vez mas de donde estabas mirando. Ahora toma el primer hueco LIBRE desde
la esquina, bajando en diagonal; si apartas la ultima, la siguiente
recupera su sitio. Extraido a utils/dropSlot con 8 tests, incluido el
caso de "la apartaron" y el tope para no salirse de la vista.
2. Que se vea. El recien anadido queda seleccionado, y la seleccion pasa de
un borde discontinuo quieto a un caminito de hormigas. El movimiento es
lo que capta el ojo en un canvas lleno; un borde fijo se pierde. Va en
un pseudo-elemento por fuera del cuerpo, sin robar clicks ni tapar el
dibujo, y se queda quieto si el sistema pide menos animacion.
3. Clicks. El izquierdo SELECCIONA y ya esta; antes abria el panel de
propiedades, o sea que no podias ni senalar una pieza sin comerte un
popup que luego habia que cerrar. Propiedades y pines pasan al click
derecho, que es donde va lo deliberado. En tactil se mantiene tocar ->
panel, que ahi no hay boton derecho.
Velxio es internacional y su galeria estaba mezclada. Tres focos:
- examples-robot-desktop: el codigo que el ejemplo ENTREGA al usuario
llevaba 33 comentarios en castellano ("Tiempo del ultimo movimiento
detectado", "Cola de estados") y 7 cadenas que el sketch imprime por
serie ("INICIO DE LA LECTURA DE SENSORES", "Movimiento DETECTADO").
Traducido todo y reescrito en ASCII, como el resto de la galeria.
- examples.ts: siete comentarios mios en castellano, de cuando anadi las
resistencias en serie a los LEDs. El resto del fichero estaba en ingles;
los deje incoherentes.
Sin cambios de comportamiento: solo texto. Los tests de galeria siguen en
verde (167).
Seam nuevo registerBoardBuiltins/getBoardBuiltins: una placa que el arbol
OSS ya dibuja (la familia Pi) puede recibir perifericos del overlay sin
registrar un ProBoardDef, que es lo que secuestraba el arte de la placa y
dejaba los cables en la esquina. El overlay lo usa para llevar los frames
del guest (cv2.imshow) al panel cableado en el canvas.
Arregla ademas el guard de connect() del bridge: comparaba readyState con
WebSocket.OPEN leido del global, asi que cuando esas constantes no estaban
`undefined === undefined` era cierto con socket a null y connect() volvia
sin abrir nada — el bridge se quedaba muerto. Ahora comprueba que el socket
exista y usa el valor numerico. Recupera los 12 tests de
multi-board-integration que esto habia roto.
El test del UART Pi->Uno pasa a comprobar el contrato vigente: por el cable
va onUartTx (el UART del header), no la consola del guest; y se anade el
caso que fija que la charla de arranque NO se filtra al vecino.
El fan-out del Interconnect envuelve el onSerialData de la INSTANCIA y
la marca con un flag. Compilar, resetear o cambiar de motor crea una
instancia nueva: sin flag y sin envoltorio, asi que la placa dejaba de
oirse por el cable a mitad de sesion. Sintoma real: la Pi en modo Linux
encendia el LED del Arduino (Pi -> Uno), pero la respuesta del Arduino
no volvia nunca (Uno -> Pi), porque el Uno habia recreado su simulador
al compilar despues de que se construyeran las rutas.
Se vuelve a enganchar en cada simulatorMap.set (siete puntos: AVR,
RP2040, RISC-V, ESP32, STM32 y los shims).
Una linea por start con el motor elegido, si estaba fijado y el motivo.
Sin ella, una placa que no arranca NADA (motor equivocado, bridge
ausente) es indistinguible de una que arranco bien: no hay error, no hay
proceso y la barra dice lo mismo. El toolbar ya deja su traza
[handleRun]; esta cubre el camino del boton de modo Linux.
startBoard llamaba `getBoardBridge(id)?.connect()`: si la placa no tenia
bridge, el encadenamiento opcional se lo tragaba y el usuario se quedaba
sin guest, sin error y con la barra mostrando Stop. Ahora avisa por
consola y baja el flag de running, que es lo unico honesto que se puede
hacer ahi.
Test nuevo del camino que usa el boton de modo Linux: fijar el modo,
parar y arrancar tiene que abrir el WebSocket del guest, dejar
engineMode en linux y running en true; y un socket en CLOSING no puede
impedir la reconexion.
connect() se rendia si el socket no estaba CLOSED, y uno en CLOSING pasa
esa prueba: pulsar "Linux terminal" justo despues de una ejecucion
cerraba el socket y el arranque siguiente no hacia nada -- ni guest, ni
error, y la barra seguia mostrando Stop. Ahora solo se rinde con OPEN o
CONNECTING y descarta el que se esta cerrando.
startBoard marca ademas running en la rama Pi: el boton de Linux
reinicia la placa por su cuenta, sin pasar por la barra, asi que el flag
se quedaba con lo que hubiera dejado la ejecucion anterior.
Las rutas serie se construyen al cargar la pagina, pero el bridge de una
placa QEMU-Linux nace al pulsar Run: ensureSerialHook encontraba bridge
nulo, hacia no-op y nadie volvia a intentarlo — los bytes que el guest
transmitia por el header salian del backend (uart_tx) y morian en un
onUartTx sin instalar. reensureSerialHooks(boardId) repite el enganche
(idempotente por el flag) y el store lo llama al crear el bridge.
Auditoria de "la Pi tiene todo lo de la placa real" con cuatro huecos
encontrados y cerrados:
1) GPIO de entrada en modo Linux: GPIO_IN respondia VAL 0 fijo (stub de
la fase 2), asi que GPIO.input() leia 0 eternamente aunque el canvas
empujara el nivel. El backend guarda ahora el ultimo nivel por pin
(set_pin_state lo escribe) y GPIO_IN contesta de ahi. Los flancos
(SET) siguen llegando al guest como antes.
2) UART del header hacia otra placa: el shim del rootfs ya hablaba
`UART <port> TX <hex>` / RX_REQ, pero sin modelo de esclavo el
backend tragaba los bytes. Ahora TX sin esclavo se emite al canvas
(uart_tx) y RX_REQ sin esclavo drena la cola que llena pi_uart_rx —
el mismo protocolo de siempre, sin ops nuevas.
3) El escaner de esclavos I2C/SPI/UART estaba doblemente muerto:
clasificaba por numero fisico de pin ('3','5','19'...) cuando el
elemento expone GPIOxx, y su unico llamador era RaspberryPiWorkspace,
que el terminal unificado reemplazo. Acepta ambos nombres y corre en
onBooted del store.
4) boardPinToNumber solo mapeaba los pines de la 3/4/5; la Zero, 1B+ y
2B (mismo header de 40 pines, mismo elemento) se quedaban sin mapa.
La placa QEMU-Linux tiene DOS flujos serie y hasta ahora el cableado
usaba el equivocado: el enrutado entregaba los bytes del vecino a la
consola (el shell) y sacaba al cable la cháchara del arranque. El
header, que es lo que el usuario cablea, no existia.
Ahora el canal de protocolo lleva dos ops nuevas:
UARTTX <b64> el guest transmitio por el header -> al canvas
UARTRX el guest pregunta que le llego -> UART_RXQ <b64>
El backend guarda una cola por instancia (acotada a 64 KB, que un script
que no lee nunca no la haga crecer) y el websocket acepta `pi_uart_rx`
con los bytes que el vecino manda. En el frontend el bridge gana
onUartTx / sendUartBytes y el Interconnect engancha ESE flujo en vez de
la consola para las placas Pi.
Con esto el mismo script -- import serial, escribir, dormir, leer --
funciona en los dos motores.
El constructor de netlist lee properties.value y cae a 1 kohm si no esta;
once ejemplos declaraban `resistance: '220'`, asi que sus LEDs lucian a
un quinto de lo previsto (brillo 0,16 en vez de 0,72 medido en el
pi-to-arduino-led-control). Se normaliza el nombre de la propiedad.
Dos piezas que faltaban para que pi-to-arduino-led-control fuera algo
mas que un guion imprimiendo lo que "habria enviado".
1) classifyPin no reconocia los pads del header por su nombre. Se llaman
GPIO14 / GPIO15 en el dibujo de la placa y en todos los cables de los
ejemplos, pero solo se aceptaban numeros fisicos: parseInt('GPIO14')
daba NaN, el pin no clasificaba como nada y el Interconnect nunca
construia la ruta. Ahora se acepta el prefijo GPIO/BCM y la numeracion
fisica sigue funcionando.
2) Seam de serie para placas que no tienen ni simulador ni bridge: el
motor de navegador corre el Python de la Pi en la propia pestana.
registerSerialSink(placa, fn) recibe los bytes que le llegan y
feedBoardSerialOut(placa, ch) anuncia los que envia, que es lo que el
enrutado por cables ya sabia repartir.
Dos fallos que salieron probando pi-to-arduino-led-control y
pi5-pir-motion-alarm.
1) BoardOnCanvas mira getProBoard() ANTES del switch OSS para decidir si
dibuja un elemento del overlay. El overlay registraba un def minimo
para las seis Pi solo para llevar una linea de setup del guest, y eso
basto para cambiarles el render: en vez de la ilustracion
Raspberry_Pi_3_illustration.svg salia la caja esquematica, con otras
coordenadas de pines, y los cables quedaban colgando en la esquina.
Ahora hay un registro aparte, registerGuestSetup(kind, linea), que
lleva la cadena y nada mas; getGuestSetup() la resuelve dando
prioridad al def del overlay si existe.
2) Las partes de entrada (PIR, botones, sensores) avisan con
simulator.setPinState(pin, nivel). En una placa QEMU-Linux no hay
simulador de MCU -- el CPU es el guest -- asi que la llamada acababa
en la instancia AVR heredada y se perdia: pulsar el sensor no hacia
nada. traceDetailed devuelve ahora tambien la placa a la que llega el
pin, y si es de la familia Pi la parte recibe un simulador que empuja
el nivel al bridge (gpio_in para el guest, el valor pin<N> que leen
los shims del motor de navegador) y al PinManager de esa placa.
Guarda de regresion para lo que se acaba de arreglar. Recorre el grafo de
cableado (no solo el vecino inmediato del LED, porque la resistencia
protege igual desde el catodo o desde el otro lado de un transistor o un
rele) y falla si algun LED llega a un pin sin limitar la corriente.
Es el peor tipo de ejemplo roto: el sketch imprime "LED ON", el runtime
quema el LED y no aparece ningun error, solo un LED oscuro.
El modelo electrico es honesto: un LED colgado directo de un pin a 3,3 V
o 5 V pide una corriente absurda, el simulador lo quema y se queda
oscuro aunque el sketch imprima "LED ON". Ya se habia arreglado en los
ejemplos de Raspberry Pi con un solo LED, pero quedaban 24 sin proteger
-- sobre todo los de Pico (i2c-scanner, spi-loopback, adc-read,
multi-protocol, serial-echo, eeprom, rtc) y todos los RGB, que necesitan
una resistencia POR CANAL.
Se anaden 36 resistencias de 220 ohm en serie, cada una entre el pin y
el anodo, colocadas al lado de su LED. Los ejemplos que ya llevaban la
resistencia en el catodo se dejan como estan: protege igual.
Con esto la galeria pasa de 24 LEDs que se queman al pulsar Play a 0.
Every Pi example hung its LED straight off a GPIO pin, so the electrical
model solved 2.4e29 A and burnt it out: the script printed LED ON while
the canvas stayed dark. Written before the SPICE engine landed, and
nobody noticed because reaching that point took a 90 s boot. 220R in
series, like the ESP32 examples already do.
Without this we cannot tell whether the in-browser path is actually
displacing guest boots — the whole point of the dual engine is a number
(% of runs that never touched the backend), so the event has to say.
A profile's extra_drive is the same file for everyone; an overlay may
need a disk built for THIS session (what the project declared). New
seam set_pi_extra_drive_resolver(fn) receives the client id, the board
and the start_pi payload and returns raw images, mounted read-only after
the profile's own. The WS route forwards msg_data and the bridge gained
startPayload so a client can declare it. Generic: no package manager,
no OS knowledge in the OSS tree.
Generic seam only — no board, OS or runtime specifics in the OSS tree:
an overlay may register an engine that decides, per board, whether a run
needs the Linux guest or can happen in the browser. The decision is data
(engine + reason + where) so the UI can explain a 90 s boot instead of
just taking it, and BoardInstance carries engineMode / enginePinned so
the user's choice is predictable and the terminal panel knows whether an
interactive shell exists. Nothing registers in OSS: the QEMU path is
untouched.
The netlist collector only understood GPIO<n>/GP<n>/bare digits, so a
wire on a QEMU-Linux board's P24 Gravity pad never got its V-source
stamped and the LED stayed dark while the guest toggled the pin.
Keeping Play enabled while the guest ran broke the convention every
other board follows (owner report). Run now disables as usual; RESET on
a booted guest re-uploads the edited files and re-runs the script
without the ~45 s reboot, matching Reset's restart-the-program meaning.
The generic arm64 image prints another product's banner/motd/login line
during boot, before guestSetup can re-brand the guest. Boards with
quietBoot show a neutral '[Velxio] Booting <label> (Linux guest)...'
progress line (dots every 4 s) while boot detection and the prompt-gated
upload still run underneath; the shell is revealed (already re-branded)
right before the auto-run command, so the user's first visible output is
their own program.
With the bare boardId as client_id, two tabs (or two users) on the same
example shared one QEMU instance: serial output went to whichever socket
connected last, keystrokes interleaved and one tab's stop killed the
other's guest. Suffix the tab session id so each tab gets its own
instance (stopped on its own disconnect, like the ESP32 workers).
QEMU-Linux boards used three competing file surfaces (workspace group
with a meaningless sketch.ino/libraries.json, the VFS panel with its
Upload button, and the Pi workspace's own editor). Now they behave like
every other board:
- the editor file group defaults to script.py for ANY Pi-family kind
(kind-based check via isPiBoardKind, not the old raspberry-pi- string)
- the libraries.json manifest row is hidden for Pi boards
- EditorPage always renders Monaco; the RaspberryPiWorkspace swap is
gone
- the bottom serial panel renders the interactive xterm (PiTerminal,
now seeded with session history) for running Pi boards
- example vfsFiles load into the editor group (single source of truth);
the run path uploads the group into the guest home
- Run on a booted guest re-runs without the 45 s reboot (Ctrl-C +
re-upload + run); starting a Pi board pops the terminal open
- piSyncAndRunScript/piRerunScript exported from the store; auto-run
now applies to the whole family (guestHome/autoRun overridable)
- ProBoardDef.autoRun: after boot (+guestSetup) the VFS uploads itself
and the command runs, so a single Run click boots, uploads and starts
the user's script (same UX as compiled boards)
- ProBoardDef.guestHome: VFS home dir override ('/root' for guests that
log in as root); those boards drop the historic hello.sh sample
- upload sequence extracted to utils/piUpload (shared by the VFS panel
button and autoRun)
- serial monitor strips DEL/C0 control echoes (backspace showed tofu)
El caso ownsPointer retornaba sin stopPropagation y el mousedown llegaba al
fondo del canvas, cuyo convenio arrastre-izquierdo-panea movia el mundo
entero bajo el dedo a mitad de swipe (solo el tap funcionaba). El modelo
tactil escucha POINTER events — stream aparte — asi que cortar el mousedown
(y el touchstart movil) no le quita nada. Los knobs wokwi conservan el
pass-through de siempre.
The boards array is replaced on every serial batch, so the attach effect
detached/re-attached built-in peripherals at up to 60 Hz while output
flowed; events landing inside the 500 ms re-attach window were lost.
One-shot streams (a guest display list) never recover — the repainting
SPI LCD decoders masked this for ESP32 boards.
isBoardSeated lee pinInfo/boardSocket del DOM; en el primer render de un
ejemplo que ABRE con la placa ya posada ni la placa ni el zocalo estan
montados, el memo devolvia 'no posada' y no recalculaba nunca (z=0 pese a
asiento exacto medido). Re-chequeo al siguiente frame y a los 400ms.
La whitelist de tags que poseen el puntero durante la simulacion gana una
salida generica regla-6a: el elemento declara `get ownsPointer() { return
true }` y el wrapper no inicia el drag mientras corre. El cristal del Round
Display pintaba el punto verde Y arrastraba el shield por el canvas a la vez.
El bump global a zIndex 3 (efecto iman del Round Display) puso TODAS las
placas por encima de TODOS los componentes: una resistencia al lado de un
Arduino quedaba oculta tras la placa en cualquier ejemplo normal (se veian
los cables, no el cuerpo). Ahora isBoardSeated (misma matematica del snap,
umbral 0.5px) decide: posada -> z3 encima de su zocalo, como el XIAO fisico
apilado en el shield; libre -> z0 debajo de los componentes, como siempre.
El iman coloca la placa exactamente en el asiento, asi que al capturarla
salta al frente y al arrancarla vuelve abajo.
- profile key extra_drive: optional read-only second virtio-blk so an
overlay can ship guest-side shim libraries (/dev/vdb)
- SENS <name> protocol op: canvas-fed named values (built-in sensors /
buttons) served from PiInstance.sensor_state, pushed by the frontend
via the new pi_sensor_state WS message
- DISP <b64> protocol op: guest display commands forwarded to the
frontend as 'display' events (built-in screens)
- RaspberryPi3Bridge: onDisplay / onGpioPwm callbacks + setSensorState
- SimulatorCanvas hands piFamily boards their Pi bridge in
attachBuiltins (was ESP32-only)
Overlay QEMU-Linux boards are not Raspberry Pis: ProBoardDef.guestSetup
lets a board send one shell line at the boot prompt (hostname/PS1/clear)
to de-brand the generic image, sent before piBooted flips so uploads
cannot interleave; the workspace start button and power-on title carry
the board's own label for non raspberry-pi kinds; the compile console
line uses the board label instead of 'Raspberry Pi 3B'.
The QEMU branch of handleRun required compiledProgram, but Pi-family
boards never have one (handleCompile early-returns 'no compilation
needed' without producing firmware), so the toolbar Run always surfaced
'Compilation produced no firmware' instead of powering the board on.
Start them directly, same as the workspace Start button.
registerPiFamilyKind + ProBoardDef.piFamily route an overlay board through
the existing Raspberry Pi bridge path (WebSocket qemu, VFS panel, boot
terminal); register_pi_board_profile lets the overlay add the matching
PI_CONFIGS entry (cpu/image_set/...) at registration time.
setAdcVoltage/setAdcWaveform solo conocian el mapa del ESP32 clasico
(GPIO32..39): en un S3 devolvian false para TODOS los pines y el part no
podia alimentar el ADC. Ahora el shim consulta bridge.adcChannelForGpio si
existe (los puentes js exponen el mapa de su chip) y cae al mapa clasico
solo si no.
El adaptador SPI del Esp32BridgeShim descartaba el MISO de los parts ('el worker
lo lleva por _spi_response'), cierto solo en la era QEMU: cualquier part SPI que
RESPONDE (una SD contestando CMD0) hablaba con nadie en modo js — medido como
SD.begin()=0 con sd_diskio reintentando CMD0 para siempre en el Round Display.
Ahora reenvia a bridge.setSpiResponse, que todos los puentes tienen: el de QEMU
lo manda al worker, y los motores js fijan el byte que su SpiForwarder devuelve
para ESTA transferencia (toda la cadena onByte corre sincrona dentro del
transfer del motor). El reposo lo restaura el decodificador compartido, que ya
completa cada byte ajeno con 0xff.
El boton conservaba animation:'none' inline fuera del estado requesting, y el
estilo inline gana a la clase: .velxio-ants quedaba con animationName none —
medido con getComputedStyle en staging. Inline solo mientras solicita permiso;
en el resto, undefined para que la clase anime.
El modulo es retrato con la botonera de pines en el borde INFERIOR y el sketch
dibuja en setRotation(1): segun el mapeo MADCTL del propio modelo (writePixel:
el borde superior del contenido rot-1 cae sobre el borde IZQUIERDO del canvas),
verlo derecho exige girar el modulo un cuarto horario — igual que en el banco
fisico. El mismo giro lleva los pines al borde izquierdo, de cara a la placa,
asi que los cables van rectos entre ambos en vez de por encima del cristal.
Cuatro tiras de gradiente repetido, una por borde, deslizando un periodo de
guion por ciclo — fondo y no border porque un borde CSS no puede animar su
dash-offset. currentColor hereda el color de estado (gris en reposo, rojo en
error). Se apaga en streaming/solicitando, y con prefers-reduced-motion.
Ojo del gato encerrado: el estilo inline usaba el shorthand background, que
pisa el background-image de la clase — pasa a backgroundColor.
Un sketch de ESP32-CAM existe para capturar, asi que esperar el click en el
toggle se lee como 'la camara esta rota': el guest arranca, esp_camera_init
completa contra el OV2640 modelado, y cam_hal agota el tiempo eternamente
esperando fotogramas que nadie envia — sin pedir siquiera el permiso.
Ahora la webcam se solicita al iniciar la ejecucion; el prompt del navegador ES
el consentimiento, y el toggle queda como apagador manual. Solo una vez por
ejecucion: parar el stream a mano no lo re-dispara.
Estaban a zIndex 0 con los componentes a 1: cualquier solape las escondia. El
caso que lo decide es el apilado — una XIAO posada en el zocalo del Round
Display debe ser lo que se ve, como la placa recien soltada queda arriba del
monton. Los montajes lado a lado no solapan, asi que nada mas cambia.
Como el iman de la breadboard pero para PLACAS: un componente puede declarar en
su elemento (estilo regla 6a, igual que pinInfo) que lleva un zocalo:
get boardSocket(): { anchorPin, accepts }
y una placa cuyo boardKind case con accepts, arrastrada cerca, se posa de golpe
con su pad anchorPin sobre el homonimo del zocalo. Un solo ancla basta porque
ambas rejillas comparten paso — esa es la gracia de un zocalo. Arrastrarla mas
alla de la tolerancia la suelta, sin estado que recordar.
Enganchado en los dos caminos de arrastre de placas (raton y tactil). El
overlay privado declara el zocalo sin que este repo sepa que existe.
Tres cosas que salieron al tirar del hilo de los pull-ups.
1) ESP32_ADC_PIN_MAP era la tabla del ESP32 CLASICO aplicada a toda la familia, y
fallaba de dos maneras:
- S3: sus pines ADC son GPIO1..20, ninguno esta en esa tabla, asi que la
busqueda devolvia undefined y el listener del potenciometro no llegaba a
engancharse. El mando no hacia nada.
- C3/C6: GPIO0 SI esta en la tabla clasica, como ADC2_CH1 -> canal 9. Pero en
esos chips GPIO0 es ADC1_CH0 y sus motores toman un indice de canal de ADC1
(0..4). El valor se empujaba al canal 9, fuera de rango, descartado en
silencio mientras el firmware leia el 0. Una respuesta equivocada en vez de
ninguna, que es peor.
adcPinMapFor(boardKind) devuelve la tabla del chip.
2) El ejemplo de las gafas OLED en MicroPython declaraba sus botones con Pin.IN a
secas. Van entre 3V3 y el pin, asi que sin pull-down el pin queda FLOTANDO con
el boton abierto. El propio ejemplo ya lo avisaba en un comentario ("Add
Pin.PULL_DOWN if the pin floats") — ahora lo hace. Es un fallo de hardware de
verdad, no un artefacto del emulador.
3) gallery-run-gate.audit.test.ts: EditorToolbar bloquea el Run cuando el
verificador de circuito saca errores, asi que un circuito invalido no es
cosmetico, es la diferencia entre un ejemplo que arranca y uno que parece
muerto. Y esos defectos se esconden hasta que el circuito resuelve DE PUNTA A
PUNTA: c3-button llevaba un LED de 506 mA que nadie veia porque sus cables
apuntaban a pines inexistentes.
El test reproduce la puerta del Run tal cual (mismo snapshot de peor caso,
mismo buildInputFromStore, mismo verifyCircuit, ngspice de verdad) sobre los
227 ejemplos. Encontro tres mas sin resistencia en serie —nano-button-led,
mega-led-chase y mega-serial-control, 17 LEDs en total— y ahora quedan 0.
A partir de aqui, un ejemplo nuevo con un LED colgado del GPIO salta en CI.
Los dos colgaban el LED directo del GPIO. Con los cables del pulsador arreglados
el circuito por fin resuelve entero, y el verificador saca lo que llevaba ahi
desde siempre: 506,9 mA por un LED de 20 mA. Como un error de circuito bloquea la
ejecucion (EditorToolbar checkOrBlock devuelve false y saca el modal de
"ejecutar de todos modos"), el ejemplo se quedaba sin arrancar.
O sea que no era un fallo nuevo, era uno viejo que hasta ahora nadie podia ver
porque el circuito no llegaba a resolverse. 220 ohm en serie, igual que el resto
de ejemplos con LED (esp32-blink-led y esp32-doom ya lo hacian bien).
El canvas reenviaba la pulsacion de un pulsador cableado como
sendPinEvent(gpio, true), o sea "pulsado = ALTO". Para el idiom canonico de
Arduino — pin -> pulsador -> GND con INPUT_PULLUP, activo a nivel BAJO, que usan
13 de los 18 ejemplos con entrada de usuario — eso esta al reves: pulsar conducia
el pin a su nivel de REPOSO y soltarlo al ACTIVO. Ademas escribia directamente en
el latch de entrada del emulador por detras de connectDigitalInputsToMcu,
desincronizando la cache lastLevel de la que ese fichero se declara unico
escritor. La victima visible era esp32-doom: cuatro botones que hacian lo
contrario de lo que pulsabas.
No lo sustituyo por la polaridad opuesta, que seria la misma adivinanza al reves:
cuando el simulador resuelve entradas por el circuito (spiceDrivenInputs), el
nivel sale del propio cableado. El INPUT_PULLUP del guest se reporta ahora como
gpio_pull, estampa una resistencia de 45k al riel de 3V3, y cerrar el pulsador
cortocircuita el nodo contra la pata que tenga al otro lado. Sale bien tanto para
un boton a GND como para uno a 3V3, sin que nadie asuma nada. El atajo se queda
solo para simuladores que se salgan del modelo electrico.
Aparte, dos arreglos de cableado:
- boardPinToNumber devuelve -1 (no null) para pines de alimentacion y masa, y
la guarda comprobaba `=== null`. Asi que la pata de GND de cada boton colaba
y registraba un SEGUNDO par de listeners apuntando al pin -1.
- c3-button y pico-button-led cableaban sus pulsadores a los pines '1a' y '1b',
que NO EXISTEN: el wokwi-pushbutton expone 1.l, 2.l, 1.r y 2.r. Esos dos
cables llevaban colgando desde siempre, asi que esos botones no han
funcionado nunca. Pasan a 1.l / 2.l, como el resto de ejemplos.
boardPinToNumber no tenia rama para 'cardputer-adv', asi que caia al return
null del final: cada cable a su cabecera EXT o al Grove Port A se conectaba a
nada. La pieza quedaba en el lienzo, el sketch leia un pin muerto y nadie
avisaba de nada.
La rama compartida de esp32 tampoco habria servido: recorta en 39 y este board
saca G40 en la cabecera. El S3 tiene 48 GPIOs.
Se anclaba a la esquina inferior derecha del lienzo y tapaba una parte del area
util del circuito, que es justo donde se trabaja. Quien pana y hace zoom no lo
necesitaba para orientarse, asi que se retira entero en vez de esconderlo tras
un interruptor mas.
Se van el componente y su CSS (no los usaba nadie mas) y la linea del listado de
novedades que lo anunciaba, que ya no seria cierta.
Las dos ramas habian divergido: master llevaba el modo lenguaje ESP-IDF puro
(#139) y v3.2 la ruta de compilacion IDF v5.5 para toda la familia ESP32 mas
los arreglos de venv/toolchain. Ambas tocaban espidf_compiler.py.
Los dos lados son ejes ORTOGONALES y se conservan enteros:
- use_idf5 / arduino_mode (v3.2): que arbol IDF usa el build (5.5 vs 4.4) y
si cabe Arduino-como-componente.
- pure_idf (master): el modo LENGUAJE que elige el usuario; sus ficheros son
las fuentes del componente main con su propio app_main().
Resolucion:
- _build_env acepta los tres. Un build IDF puro fuerza arduino_mode a falso:
la plantilla CMake mete el componente arduino-esp32 en cuanto existe
ARDUINO_ESP32_PATH, asi que dejarlo puesto compilaba el core de Arduino en
un build que no tiene sketch. Lo cazaron los tests de master.
- VELXIO_PURE_SKETCH solo con pure_idf, nunca con arduino_mode a falso a
secas: un target sin core arduino-esp32 sigue entregando un SKETCH al
traductor legacy y no debe tomar la rama del glob puro.
- La identidad del build-dir suma los dos tokens (|idf:N|ard:N y |lang:pure):
ningun par de esas combinaciones puede compartir un build/ configurado.
- La cadena de escritura de fuentes queda pure_idf -> arduino_mode -> legacy.
- sdkconfig: render de v3.2 (con target/use_idf5) mas el filtrado de simbolos
CONFIG_ARDUINO* de master cuando el build es puro.
test/backend/unit/test_espidf_compiler.py: los 7 tests que ya estaban rotos en
v3.2 (AttributeError: idf5_path, fixture sin actualizar desde que se anadio la
seleccion de IDF) vuelven a pasar.
Verificado: backend 293 pasan / 0 fallan (v3.2 traia 7 rotos); frontend 2268
pasan / 0 fallan en los dos shards.
connectMcuEdgesToService resubscribes its per-pin listeners whenever
pinNetMap changes. The subscription swept pin NUMBERS 0..63 and
reverse-mapped them to names ('GPIO2' on ESP32, 'GPIO17' on Pi) to match
against pinNetMap's keys — but those keys are the WIRE pin names ('2',
'4', 'A0'), so after the first solve the match failed for every pin and
the resubscription attached nothing. Any mid-run pinNetMap update then
silently killed the MCU-edge → SPICE path and the canvas froze at the
last solved state while the firmware kept toggling.
Masked until now because nothing perturbed pinNetMap mid-run on the
blink examples; pure ESP-IDF mode (#139) unmasked it — gpio_reset_pin()
leaves the internal pull-up enabled, the worker reports gpio_pull, the
handler requests an electrical resolve, pinNetMap gets a new identity,
and the ESP-IDF blink example's LED froze ON.
Fix: subscribe FROM the pinNetMap names, mapped to PinManager pins with
the same pinNameToArduinoPin the netlist collector uses (STM32 via
stm32PinNameToLinear), and hand schedulePin the netlist name so
handleMcuEdge's v_<board>_<pin> lookup hits the fast alterSource path
instead of a full rebuild per edge. The 0..63 sweep remains as the
pre-first-solve fallback. Also fixed pinNameToArduinoPin's dead 'GPIO'
branch ('GP' tested first turned 'GPIO32' into parseInt('IO32') = NaN).
Adds a third entry to the board language selector next to Arduino C++
and MicroPython: ESP-IDF. In this mode the user writes a plain ESP-IDF
project — app_main() entry point, FreeRTOS + driver APIs — and the
backend compiles it through the same ESP-IDF toolchain it already uses
for ESP32 Arduino sketches, just without the arduino-esp32 component.
Backend:
- CompileRequest.language ('espidf') threaded through the sync + async
compile paths and folded into the dedup job key (language='arduino'
and omitted hash identically so old clients keep dedupping).
- espidf_compiler: pure_idf flag. User files are written into main/
as-is (no Arduino.h wrap, no velxio_compat.h, Arduino library
resolution skipped), ARDUINO_ESP32_PATH is dropped from the build env
and VELXIO_PURE_SKETCH raised so the template CMake compiles the
user's own sources via a glob branch. Pure builds get their own
persistent build-dir variant through the eff_hash fold.
- QEMU WiFi compat for IDF-style code: esp_wifi.h/esp_wifi_init
detection sets has_wifi, and literal #define SSID/PASS plus
wifi_config_t designated initializers are normalized to the QEMU AP.
- CONFIG_ARDUINO_* lines are stripped from sdkconfig.defaults in pure
mode (the symbols don't exist without the arduino component).
Frontend:
- LanguageMode gains 'espidf'; BOARD_SUPPORTS_ESPIDF covers the ESP32
family (Xtensa, S3, C3). Toolbar shows the option only for those.
- Switching modes seeds a main.c blink skeleton (app_main + gpio
driver), mirroring the MicroPython main.py flow.
- compileCode sends language='espidf'; run/stop paths are unchanged
(the QEMU worker consumes the same merged flash image).
- New gallery example: esp32-idf-blink (LED + resistor on GPIO 2).
Tests: unit coverage for the build-env switch, IDF wifi normalization,
job-key variance, file-group seeding and the new example; verified
end-to-end in a container from the prod image (pure build produces a
bootable flash image; Arduino-mode build unchanged, same variant hash).
componentToSpice/isSpiceMapped now consult a pro-registered mapper table after
the static MAPPERS, so a private build can give its closed components a SPICE
model (e.g. the DFRobot Gravity analog sensors emit a voltage source at AOUT).
Empty in a pure OSS build. Mapper type exported.
Round Display, Grove Gesture and ReSpeaker Lite now advertise themselves in
the OSS picker like the M5Stack matrices; they auto-hide in any build whose
overlay merges the real components (registry.getById).
A wire to a pro board's Dx pad (XIAO 'D2', etc.) did nothing because
isBoardComponent only knew the OSS board-id list, so the electrical/sensor
chain skipped the board endpoint. It now also returns true for any
proBoardRegistry kind, so boardPinToNumber (which already consults the pro
def's pinToNumber) resolves the Dx name. Fixes the numeric-alias-only
workaround in the XIAO examples.
registerSensorControls() lets a private build add SensorControlDef entries for
sensors it ships outside the OSS tree (e.g. the DFRobot Gravity analog family)
so they get the live slider panel; every SENSOR_CONTROLS lookup now goes
through getSensorControl(id) which falls back to the registered map. Dead code
in a pure OSS build, same contract as proBoardRegistry / registerComponentDoc.
xiao-esp32s3-sense / xiao-esp32c6 / xiao-rp2040 advertise themselves in the
OSS picker like the other hosted-only boards; the ads auto-hide in any build
whose overlay registers the real kinds (same BOARD_KIND_LABELS contract).
filteredComponents' new registryVersion dep evaluated in its useMemo deps
array while the const was still in the temporal dead zone (declared further
down the component) — 'Cannot access N before initialization', white screen
on every page. Hooks moved up next to the registry declaration.
The @pro overlay import is dynamic, so board/component registration can land
AFTER the picker mounted and memoized its lists. allBoards had frozen deps —
if the picker rendered first, overlay boards (M5Stack Core, Cardputer,
Pimoroni, C6) vanished for the whole session while their ONLINE ads were
already hidden; the component grid + ONLINE component ads flipped between ad
cards and real entries depending on who won the reload race (different SVG,
ONLINE badge appearing and disappearing, wrong hover thumbnail).
proBoardRegistry and ComponentRegistry.mergeComponents now bump a version and
notify subscribers; the picker's allBoards / filteredComponents /
visibleComponentAds memos key off those via useSyncExternalStore — same
contract as proRoutes and registerProExamples.
- The card preview creates the live element but only forwarded
defaultValues.value, so variants sharing a tag rendered identically (both
M5Stack Chain matrices showed the dark RGB housing — the mono flag never
reached the element). Forward every defaultValues entry, matching what
DynamicComponent assigns at placement.
- The hover datasheet panel sat at z-index 2000 while the picker overlay was
raised to 9000 (above the AI chat), hiding the popover behind the very
modal that summons it. Raise it to 9100.
Replaces the static hero-editor screenshot with the animated
estacion-meteorologica-esp32.gif (whole circuit executing). contain +
matching background so the full circuit stays visible instead of
cover-cropping it.
Boards with built-in hardware (LCD on the element's own canvas, speaker,
on-board buttons/keyboard) need run-time wiring between the DOM element and
the board's simulator shim / ESP32 bridge. One generic SimulatorCanvas effect
now hands those handles to the overlay's attachBuiltins shortly after run
start and runs its cleanup on stop — no board names in the OSS tree.
registerProExamples() now notifies useSyncExternalStore subscribers (same
contract as proRoutes): a direct URL to an overlay-registered example resolved
before the dynamic @pro import landed and stuck on the 404 branch. The three
example pages subscribe to the version counter.
registerProBoards() lets a hosted overlay ship boards outside the OSS tree as
data-only definitions: registration patches the exported BoardKind maps
(labels / FQBN / MicroPython) so every existing read site keeps working, and
the sites a map can't cover consult the registry — canvas render (custom
element or overlay render fn), BOARD_SIZE, pin-name mapping, picker list +
descriptions + tag, ESP32 family routing, in-browser simulator construction
and firmware load (structural ProBoardSimulator contract, duck-typed PIO
attach/detach), built-in bridge sensors, and a CS-gated built-in microSD
(sdCsPin -> sd_card.cs_pin worker config). Esp32Bridge additionally gains the
esp32-c6 machine type + TX pin (public chip knowledge — the C6 compile path
already ships) and a generic sendKey() for built-in matrix keyboards.
registerProExamples() appends gallery examples at runtime; the board ONLINE
ads recompute at render so registration hides them. OSS behavior without an
overlay is unchanged — the registry is dead code, same as the other seams.
Components that only exist in the hosted editor now advertise themselves in
the picker exactly like the online-only boards do: ONLINE_ONLY_COMPONENT_ADS
(same module as the board ads) renders an ONLINE-badged card that links to
velxio.com, auto-hidden in any build whose ComponentRegistry has the real
component (the hosted overlay merges it in — no per-build switches). First
entries: the M5Stack Chain RGB/Mono 8x8 matrices. componentDocs gains
registerComponentDoc(id, raw) so an overlay can supply the hover datasheet
for components it injects at runtime.
Gallery templates and agent-loaded projects can store element tag names
("wokwi-lcd2004") where the registry keys by the bare id ("lcd2004").
getById now falls back to the stripped id, so such a component renders
instead of sitting invisible in the store — a real agent session shipped
an LCD counter whose LCD existed in the store, was wired and validated,
and simply never appeared on the canvas.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Root cause of "el agente construye el reloj, dice que funciona, pero el
display queda en blanco hasta recargar la página" — diagnosed by driving
the live agent end-to-end and instrumenting the element:
The 7-segment part simulator caches its state (segments, digitValues,
digitEnabled) in a WeakMap keyed by the DOM element, sizing it from
element.digits at FIRST access. The agent builds incrementally: it adds
the display with the default digits=1 and only then sets digits=4 — so
the cached state was born in single-digit mode. Every later attachEvents
(compile bumps hexEpoch → re-attach with the finished wiring) kept
consulting the stale state: it subscribed COM.1/COM.2 (which don't exist
on a 4-digit part) instead of DIG1..DIG4, and because those resolvers DID
attach, the all-digits-on fallback never kicked in either. Result: no
digit ever enabled, no flush ever ran, values stayed a frozen 8-zero
array. A page reload "fixed" it because the fresh element mounted with
digits already 4.
get7SegState now compares the cached digit count against the element's
current value and rebuilds the state when they differ, so any re-attach
after a digits change subscribes the right pins.
Test: attach with digits=1 (COM subscribed), set digits=4, re-attach →
DIG1..4 subscribed, and a segment+digit pulse actually lights values[0]
in the 32-slot array.
Verified live: the exact agent prompt that produced a permanently blank
display now shows the multiplexed digits + blinking colon in-session,
no reload.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The stop-first guard in the Run button only fires when board.running is
true, but the Run button is DISABLED while a board runs — so by the time the
user can actually click Run, the board has already disconnected
(running=false) and the guard is a no-op. The failure lives one level down:
Esp32Bridge.connect() early-returned whenever a socket lingered in ANY
non-CLOSED state (CONNECTING/OPEN/CLOSING). The agent's run_simulation
leaves such a socket; when its backend QEMU session ends but the frontend
socket is still zombie, the user's Run → startBoard → connect() did nothing.
A page reload "fixed" it only by constructing a fresh bridge.
connect() now tears down any lingering socket (detaching handlers + close)
and opens a new one to the same session key — exactly what the reload does,
which is why the reload always worked. The backend already handles a new WS
replacing an existing session, so no reload is needed.
Test: connect() on an OPEN socket closes the old one and boots a fresh
start_esp32 (esp32-dht22-flow).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two issues from a real ESP32 7-segment clock the agent built.
Run after the agent didn't work until a page reload
---------------------------------------------------
The agent's run_simulation leaves the ESP32 board RUNNING (live QEMU
WebSocket). Esp32Bridge.connect() is a no-op while the socket is non-CLOSED,
so the user's subsequent Run click called startBoard() → connect() → did
NOTHING. And if the backend QEMU session had since died while the frontend
socket lingered (CONNECTING/OPEN/CLOSING), the user saw a dead sim that only
a reload cleared — exactly the "di Run y no funcionó; recargué y sí" report.
The Arduino/C++ QEMU path now stops a running board first (closing the WS),
waits for it to settle, then boots fresh — the MicroPython path already did
this for the same reason.
Wires painted over the 7-segment digits
----------------------------------------
The agent bridges each segment strip to its resistor from a breadboard hole
that is physically UNDER the seated display; on the flat canvas those wires
(wire layer z 35) painted over the digits (component z 1) — "casi ni se ven
los dígitos". A large-bodied display seated on a breadboard now renders
ABOVE the wire layer, so its face occludes the wires crossing it exactly as
the real part's body would (the wire passes behind it to reach the hole).
Scoped to display bodies (7segment, matrix, oled, lcd, ili9341, led-ring…)
and only when actually seated; thin parts and free-floating displays are
untouched. The pin overlay + seated-pin markers share the display's stacking
group, so they rise with it and wiring still works.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fixes the reported breadboard wiring UX ("requerimos un vocabulario"):
Vocabulary implemented (breadboardOccupancy.ts, pure + unit-tested):
- 1 hole = 1 wire. A hole already holding a visible wire can't start a
new one — clicking it SELECTS that wire. This is the core fix: wires
running hole-to-hole across the board were impossible to select
because the pin overlays swallowed every click and silently started a
new wire (so the top horizontal rail wire was un-deletable).
- Same 5-hole strip / rail = one net. When a new wire end lands in an
occupied hole (a seated leg or another wire), it shifts to the
NEAREST FREE hole of the same group — electrically identical, the
real-world "bridge to the next hole in the row". Never crosses strips.
Two selection bugs behind the symptom:
- Click on a wire lying over the breadboard BODY now selects the wire
instead of opening the breadboard's 830-hole property dialog (that
list popping over everything was the "se sobrepone la lista de todos
los puntos" report). Guarded so the bubbled canvas click doesn't
re-toggle the fresh selection.
- Click on a hole occupied by a wire selects the wire (handlePinClick),
so wires anchored in holes are reachable at all.
Jumper colors (like a real kit — a board of identical green wires is
unreadable, "se ven todos verdes"):
- Power-rail holes mandate red (tp./bp. = +) / black (tn./bn. = −).
- Other breadboard holes get a random jumper-palette color on manual
draw; red and black are reserved for rails.
- jumperColorForId gives agent/deterministic callers a stable per-wire
color across reloads.
Tests: breadboard-occupancy.test.ts (12) — findWireAtHole (skips seating
wires, topmost wins), resolveFreeHole (same-strip shift, no cross-strip,
rail shift, passthrough), color policy (rails, palette determinism,
red/black reserved).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
David Montero